JP7045243B2 - Computer program, welding information calculator, welding torch, welding power supply, welding system - Google Patents

Description

The present invention relates to a computer program, a welding information calculation device, a welding torch, a welding power supply, and a welding system.

The consumable electrode type arc welding system is equipped with a welding power supply, a wire feeder and a welding torch. The welding wire is fed to the welding torch by a wire feeding device, and the welding power supply supplies power to the welding torch to generate an arc between the welding wire and the work to weld the work. In a semi-automatic arc welding system, a welding operator grips the welding torch and brings the tip of the welding torch close to the work piece to be welded to perform welding.

Generally, the optimum welding current, welding voltage, etc. are determined according to the material of the work, the material of the welding wire, the diameter, and the like. However, when the moving speed and posture of the welding torch change, the optimum welding current and the like also change. For this reason, at a site where the welding posture changes depending on the welding location, the welding operator needs to appropriately adjust the welding current and the like each time the welding posture changes. Adjustment of welding current and the like is often performed based on the experience and intuition of the welding worker, and improvement of welding quality and workability largely depends on the skill level of the welding worker.
In addition, it is difficult to keep the moving speed and posture of the welding torch constant, and skill is required. When the moving speed and the posture of the welding torch change, the optimum welding current and the like change, and as a result, the welding quality deteriorates.

Patent Document 1 discloses a technique of storing a correction value such as a welding current according to a change in the angle of a welding torch and automatically correcting the welding current or the like according to a change in the posture of the welding torch. According to Patent Document 1, it is not necessary to manually adjust the welding current or the like according to the change in the posture of the welding torch, and the welding quality and workability can be improved.

Japanese Unexamined Patent Publication No. 2013-66999

However, in reality, it is difficult to store in advance the correction values such as the welding current according to all changes such as the moving speed of the welding torch, the posture of the welding torch with respect to the moving direction, and the angle of the nozzle of the welding torch with respect to the vertical direction. Is. When the combination of the posture of the welding torch and the correction value registered in the table is limited to a certain amount, there is a problem that the welding current or the like cannot be sufficiently corrected and the optimum welding current or the like cannot be obtained.

Further, other types of welding such as laser welding, electron beam welding, and gas welding also have the same problems as described above.

An object of the present invention is to adapt to changes in the moving speed of the welding torch, the posture of the welding torch with respect to the moving direction, the angle of the nozzle of the welding torch with respect to the vertical direction, etc., without relying on the skill level of the welder in semi-automatic welding. To provide a welding information calculation device that can calculate welding information such as the optimum welding current and welding voltage, a welding torch, a welding power supply and a welding system, and a computer program for learning the welding information calculation device. be.

The computer program according to this embodiment tells the computer speed information indicating the speed of the welding torch for welding, attitude information indicating the attitude of the welding torch with respect to the moving direction, and angle information indicating the angle of the welding torch with respect to the vertical direction. At least one is acquired, and based on at least one of the acquired speed information, attitude information, and angle information, the welding current or welding voltage for performing welding adapted to the moving speed, attitude, and angle of the welding torch is included. Information for determining the quality of the welding result based on the welding information calculated by the welding information calculation unit or the quality of the welding result based on the information so that the welding information calculation unit that calculates the welding information can be machine-learned. A process of acquiring the judgment result shown and storing at least one of the acquired speed information, attitude information, and angle information in the storage unit in association with the information for judging the quality of the welding result or the judgment result of the quality is executed. It is a program to make it.

The welding information calculation unit is a circuit or functional unit that calculates welding information adapted to the moving speed, posture, and angle of the welding torch. Welding information is information for controlling a welding power source, and is, for example, information such as a welding current, a welding voltage, and a welding speed of a welding wire. The welding information may be information indicating a correction amount based on the standard welding information at a predetermined moving speed, a predetermined posture and an angle of the welding torch.
The computer program according to this embodiment causes the computer to execute a process of collecting and accumulating learning information necessary for machine learning of the welding information calculation unit. Specifically, the computer collects the speed information, the attitude information, and the angle information of the welding torch, and the judgment result indicating the quality of the welding result using the welding information calculated by the welding information calculation unit based on the information. .. Then, the computer stores the collected speed information, attitude information, angle information, and quality determination result in the storage unit as learning information of the welding information calculation unit.
Further, instead of the determination result indicating the quality of the welding result, information for determining the quality of the welding result may be collected and stored in the storage unit. The information for determining the quality of the welding result is, for example, information such as a welding current, a welding voltage, and a welding sound detected during welding.
As described above, according to this aspect, welding information adapted to the moving speed, posture and angle of the welding torch can be calculated, and information that makes it possible to further improve the welding result can be collected and accumulated.

The computer program according to this aspect is a program for causing the computer to execute a process of machine learning the welding information calculation unit so that the welding result is improved based on the information stored in the storage unit. ..

The computer program according to this embodiment causes the computer to execute a process of machine learning the welding information calculation unit using the collected and accumulated information. Welding results can be improved by machine learning the welding information calculation unit. That is, the content of the welding information calculated when specific speed information, attitude information, and angle information are input to the welding information calculation unit is gradually corrected so that welding information that can obtain better welding results can be calculated. Become.

In the computer program according to this aspect, the information for determining the quality of the welding result is the welding current and welding voltage detected during welding, the feeding speed of the welding wire, the short circuit condition, and the welding sound collected during welding. It is a program for causing the computer to execute a process of determining the quality of a welding result based on the acquired information, including information indicating sound and at least one image of a welded portion captured after welding.

According to this aspect, the quality of the welding result can be automatically determined, and the welding information calculation unit can be machine-learned. For example, the computer has information such as the welding current and welding voltage detected during welding, the feed rate of the welding wire, the short circuit condition, the welding sound collected during welding, and the image of the welded portion imaged after welding. Is used to determine whether the welding result is good or bad, and the welding information calculation unit is machine-learned.

The welding information calculation device according to this embodiment has at least speed information indicating the speed of the welding torch for welding, attitude information indicating the attitude of the welding torch with respect to the moving direction, and angle information indicating the angle of the welding torch with respect to the vertical direction. To perform welding adapted to the moving speed, attitude and angle of the welding torch based on at least one of the acquisition unit to acquire one and the speed information, attitude information and angle information acquired by the acquisition unit. It is provided with a learned welding information calculation unit that calculates welding information including welding current or welding voltage.

The welding information calculation device of this embodiment acquires speed information, attitude information and angle information of the welding torch, and the welding information calculation unit calculates welding information adapted to the moving speed, attitude and angle of the welding torch. The welding information calculation unit is a learned circuit or functional unit that can calculate the welding information. The contents of the welding information are as described above.
A welding system equipped with a welding information calculation device according to this embodiment can perform welding control using welding information adapted to the moving speed, posture and angle of the welding torch, and can perform welding control of the moving speed, posture and angle of the welding torch. Welding quality can be maintained regardless of changes.

The welding information calculation device according to this embodiment includes a welding condition information acquisition unit that acquires welding condition information indicating welding conditions that serve as a reference for welding information at a predetermined speed, attitude, and angle of the welding torch, and calculates the welding information. Based on at least one of the speed information, attitude information, and angle information acquired by the acquisition unit and the welding condition information acquired by the welding condition information acquisition unit, the unit determines the state and welding conditions of the welding torch. Calculate adapted welding information.

According to this aspect, the welding information calculation unit is based on the welding condition information which is the reference of the welding information at a predetermined speed, posture and angle of the welding torch, and at least one of the speed information, the posture information and the angle information. Welding information suitable for the state of the welding torch and welding conditions is calculated.
Welding condition information includes the welding current, welding voltage, and welding wire recommended for the moving speed, posture, and angle of a predetermined welding torch, such as the material, thickness, groove shape, joint type, and work posture of the work. This is information that serves as a reference for determining the feeding speed, etc. Further, the welding condition information may be information indicating a welding current, a welding voltage, a welding wire feeding speed, etc. as standard or reference values.

In the welding information calculation device according to this embodiment, when at least one of speed information, attitude information and angle information is input, the welding information calculation unit obtains welding information adapted to the moving speed, attitude or angle of the welding torch. It has a neural network trained to output.

According to this aspect, the welding information calculation unit uses the trained neural network to input welding information adapted to the moving speed and attitude of the welding torch when the speed information, attitude information and angle information of the welding torch are input. Can be output.

In the welding information calculation device according to this aspect, the welding information calculation unit includes a plurality of different neural networks according to welding conditions.

According to this aspect, the welding information calculation unit includes a plurality of trained neural networks specialized for a plurality of welding conditions. By specializing each neural network for specific welding conditions and learning, it is possible to suppress the information required for machine learning and to learn efficiently.

In the welding information calculation device according to this aspect, when the welding information calculation unit inputs at least one of speed information, attitude information and angle information and welding condition information, the moving speed, attitude or angle of the welding torch is input. It is equipped with a neural network trained to output welding information adapted to.

According to this aspect, the welded information calculation unit adapts the trained neural network to the welding conditions and the moving speed and attitude of the welding torch when the welding conditions, the speed information, the attitude information and the angle information of the welding torch are input. Welding information can be output.

The welding information calculation device according to this embodiment has a welding state information acquisition unit that acquires information for determining the quality of the welding result performed based on the welding information calculated by the welding information calculation unit, and the welding state information. A pass / fail judgment unit for determining the quality of the welding result based on the welding result information acquired by the acquisition unit is provided, and at least one of the speed information, the attitude information, and the angle information acquired by the acquisition unit and the quality determination. A learning processing unit for machine-learning the welding information calculation unit is provided so that the welding result is improved based on the determination result of the unit.

According to this aspect, the welding information calculation device acquires information for determining the quality of the welding result or the information, and the quality determination unit determines the quality of the welding result. The information for determining the quality of the welding result was captured during welding, the welding current and the welding voltage, the feeding speed of the welding wire, the short circuit condition, the welding sound collected during welding, and the image after welding. Information such as images of welded parts. Then, the learning processing unit causes the welding information calculation unit to perform machine learning based on at least one of the speed information, the attitude information, and the angle information of the welding torch and the judgment result of the quality of the welding result.
Therefore, the welding information calculation unit can be automatically machine-learned so that the welding result is improved.

In the welding information calculation device according to this aspect, the information for determining the quality of the welding result is collected in the welding current and welding voltage detected during welding, the feeding speed of the welding wire, the short circuit condition, and the sound collected during welding. Includes welding noise and information indicating at least one image of the weld site imaged after welding.

According to this aspect, the quality of the welding result is judged based on the welding current during welding, the welding voltage, the feeding speed of the welding wire, the short circuit condition, the welding sound, the welding image, etc., so that the welding result is improved. , The correction amount calculation unit can be machine-learned.

In the welding information calculation device according to this aspect, the welding information calculation unit has the moving speed and attitude of the welding torch based on at least one time change of the speed information, the attitude information and the angle information acquired by the acquisition unit. Alternatively, it has a prediction unit that predicts the angle, and adapts to the state of the welding torch based on at least one of the speed information, the attitude information, and the angle information indicating the moving speed, the posture, and the angle predicted by the prediction unit. Welding information is calculated.

According to this aspect, the welding quality can be improved more effectively by predicting the time change of the moving speed, the posture and the angle of the welding torch and calculating the welding information adapted to the prediction result.

In the welding information calculation device according to this aspect, when at least one of speed information, attitude information and angle information is input in time series, the prediction unit determines the predicted movement speed, attitude or angle of the welding torch. A plurality of neural networks that output the indicated speed information, attitude information, or angle information, and a reception unit that accepts the selection of the neural network are provided, and the prediction unit uses the neural network received by the reception unit. , Predict the moving speed, posture or angle of the welding torch.

According to this aspect, the prediction unit includes a plurality of neural networks for predicting the moving speed, posture or angle of the welding torch. Welders can choose which neural network to use. That is, it is possible to select a neural network that has learned the habits of each welder. The welding information calculation device accepts the selection of the neural network at the reception unit, predicts how to move the welding torch differently for each welding operator, and welds information based on the predicted moving speed, posture and angle of the welding torch. Is calculated.

The welding torch according to this embodiment is a welding torch provided with any one of the above welding information calculation devices and is supplied with power from a welding power source to perform welding, and the welding information calculation device is provided on the welding torch. It is provided with a transmission unit that transmits the welding information calculated by the welding information calculation unit to the welding power source.

According to this aspect, the welding torch is provided with a welding information calculation device. The welding information calculation device provided in the welding torch calculates welding information adapted to the moving speed, posture and angle of the welding torch, and outputs the calculated welding information to the welding power supply. By giving the welding information, the welding current, welding voltage, etc. adapted to the state of the welding torch can be supplied from the welding power source to the welding torch.

The welding power source according to this aspect is a welding power source that includes any one of the above-mentioned welding information calculation devices and supplies electric power to the welding torch, and the acquisition unit is a speed information and an attitude transmitted from the welding torch. At least one of the information and the angle information is acquired, and the welding power supply supplies electric power based on the welding information calculated by the welding information calculation unit.

According to this aspect, the welding power source includes a welding information calculation device. The welding information calculation device provided in the welding power supply calculates welding information adapted to the moving speed, attitude and angle of the welding torch, and based on the calculated welding information, the welding current and welding adapted to the state of the welding torch. Voltage and the like can be supplied to the welding torch.

The welding system according to this aspect is a welding system including any one of the above welding information calculation devices, a welding power source that supplies power to the welding torch, and a welding torch that is supplied with power from the welding power source to perform welding. The acquisition unit acquires at least one of speed information, attitude information, and angle information transmitted from the welding torch, and the welding power supply is calculated by the welding information calculation unit. Power is supplied based on the welding information.

According to this aspect, a welding information calculation device is provided at an arbitrary position in a welding system equipped with a welding power supply and a welding torch. The welding power supply can supply the welding current, welding voltage, etc. adapted to the state of the welding torch to the welding torch based on the welding information calculated by the welding information calculation device.

According to this aspect, in semi-automatic welding, it is possible to adapt to changes in the moving speed of the welding torch, the posture of the welding torch with respect to the moving direction, the angle of the nozzle of the welding torch with respect to the vertical direction, etc. without relying on the skill level of the welding operator. Welding information such as the optimum welding current and welding voltage can be calculated.

It is a schematic diagram which shows the arc welding system which concerns on Embodiment 1. FIG. It is a block diagram which shows the structural example of the welding torch which concerns on Embodiment 1. FIG. It is a block diagram which shows the structural example of the welding power source which concerns on Embodiment 1. FIG. It is an external view which shows the structural example of a welding torch. It is a schematic diagram which shows the relationship between the moving speed and welding voltage of a welding torch, and the welding state. It is a schematic diagram which shows the posture of the welding torch with respect to the moving direction. It is a schematic diagram which shows the angle of the welding torch with respect to the vertical direction. It is a block diagram which shows the structural example of a learning part. It is a block diagram which shows the structural example of the welding torch which concerns on Embodiment 2. It is a block diagram which shows the structural example of the welding torch which concerns on Embodiment 3. It is a block diagram which shows the structural example of the quality determination part which concerns on Embodiment 3. It is a block diagram which shows the structural example of the learning part which concerns on Embodiment 4. It is a block diagram which shows the structural example of the learning part which concerns on Embodiment 5. It is a block diagram which shows the structural example of the learning part which concerns on Embodiment 6.

Hereinafter, the present invention will be described in detail with reference to the drawings showing the embodiments thereof. In addition, at least a part of the embodiments described below may be arbitrarily combined.
(Embodiment 1)
FIG. 1 is a schematic view showing an arc welding system according to the first embodiment. The consumable electrode type arc welding system (welding system) includes a welding power supply 1, a wire feeding device 2, and a welding torch 3. The arc welding system is semi-automatic.

<Welding power supply>
The welding power supply 1 includes a first output terminal and a second output terminal for supplying electric power to the welding torch 3, and a signal terminal for transmitting and receiving signals. One end of the first power cable 41 is connected to the first output terminal of the welding power supply 1, and the other end of the first power cable 41 is connected to the welding torch 3 via the wire feeding device 2. The second output terminal of the welding power supply 1 is connected to the work W by the second power cable 42. The work W is grounded. The welding power supply 1 converts the three-phase alternating current of the power system P into the required welding current and welding voltage, and supplies the power required for arc welding to the welding torch 3 through the first and second power cables 41 and 42.

The welding power supply 1 and the wire feeding device 2 are connected by a power transmission line 5 for drive control. Further, the wire feeding device 2 and the welding torch 3 are also connected by the power transmission line 5. The welding power supply 1 supplies electric power for driving the feeding motor of the wire feeding device 2, the control unit 36 of the welding torch 3 (see FIG. 2), and the like through the power transmission line 5, the wire feeding device 2 and the welding torch 3. Supply to.

The arc welding system includes a gas cylinder 6 that supplies a shield gas for preventing oxidation of the molten metal during arc welding. One end of the gas pipe 7 is connected to the gas cylinder 6, and the other end of the gas pipe 7 is connected to the welding torch 3 via the welding power supply 1 and the wire feeding device 2. The shield gas of the gas cylinder 6 is supplied to the welding torch 3 through the gas pipe 7.

One end of the signal line 8 is connected to the signal terminal of the welding power supply 1, and the other end of the signal line 8 is connected to the wire feeding device 2. Further, the wire feeding device 2 and the welding torch 3 are also connected by a signal line 8. The welding power supply 1 outputs a control signal to the wire feeding device 2 through the signal line 8 to control the feeding speed of the welding wire and the like. Further, the welding power supply 1 and the welding torch 3 are in the state of the welding torch 3, such as the moving speed of the welding torch 3, the attitude of the welding torch 3 with respect to the moving direction, the angle of the nozzle 371 (see FIG. 4) of the welding torch 3 with respect to the vertical direction, and the like. Sends and receives various information necessary for welding control adapted to.

<Wire feeder>
The wire feeding device 2 is a device that feeds a welding wire that functions as a consumable electrode to the welding torch 3. The welding wire is, for example, a solid wire. The wire feeding device 2 and the welding torch 3 are connected by a torch cable 39, and the welding wire passes through the inside of the liner provided inside the torch cable 39 and the welding torch 3 and the tip of the welding torch 3. Guided by. The wire feeding device 2 includes a feeding roller for feeding the welding wire, a feeding motor, and the like, and is driven by the electric power supplied from the welding power source 1.
Further, inside the torch cable 39, a first power cable 41, a gas pipe 7, a liner, a power transmission line 5, and a signal line 8 are arranged. The drive control electric power supplied from the welding power source 1 to the wire feeding device 2 is also supplied to the welding torch 3 through the power transmission line 5 arranged inside the torch cable 39. The wire feeding device 2 can communicate with the welding power source 1 and the welding torch 3 through the signal line 8. Similarly, the welding power supply 1 and the welding torch 3 can communicate with each other via the wire feeding device 2.

<Welding torch>
The welding torch 3 is made of a conductive material such as a copper alloy, and has a cylindrical contact tip that guides the welding wire to the work W to be welded and supplies the welding current required for generating an arc. The welded wire fed from the wire feeding device 2 is fed so as to protrude from the tip of the contact tip. The first power cable 41 is electrically connected to the contact chip. The contact tip contacts the welding wire penetrating the inside thereof, and a welding current is supplied to the welding wire.
Further, the welding torch 3 has a hollow cylindrical shape surrounding the contact tip, and has a nozzle 371 for injecting a shield gas from the opening at the tip to the work W. The shield gas is for preventing oxidation of the work W and the welding wire melted by the arc. The shield gas is, for example, a carbon dioxide gas, a mixed gas of carbon dioxide gas and an argon gas, an inert gas such as argon, and the like.

FIG. 2 is a block diagram showing a configuration example of the welding torch 3 according to the first embodiment. The welding torch 3 includes a torch-side communication unit 31, a display unit 32, an operation unit 33, a storage unit 34, a torch state detection unit 35, and a control unit 36.

The torch-side communication unit 31 is a circuit that communicates with the wire feeding device 2 or the welding power supply 1. The torch-side communication unit 31 modulates the signal given from the control unit 36 according to a predetermined communication protocol, and transmits the signal to the wire feeding device 2 or the welding power supply 1 through the signal line 8. The communication protocol is, for example, CAN (Controller Area Network). Further, the signal transmitted from the welding power supply 1 and the wire feeding device 2 is received, demodulated, and the demodulated signal is given to the control unit 36. The communication between the welding torch 3 and the welding power supply 1 is not limited to the wired communication, and may be wireless communication.

The display unit 32 has a display 321 (see FIG. 4) that displays various information related to welding. The display 321 is, for example, a liquid crystal display panel. For example, the material, thickness, groove shape, joint type, welding conditions such as the posture of the work W, recommended welding current, welding voltage, moving speed of the welding torch 3, posture, angle with respect to the vertical direction, etc. Is displayed.

The operation unit 33 is various switches and buttons for accepting operations by the welding operator. When the operation unit 33 is operated, an operation signal is input to the control unit 36, and the control unit 36 can recognize the operation state of the operation unit 33.

The storage unit 34 is a non-volatile memory such as an EEPROM (Electrically Erasable Programmable ROM) or a flash memory. The storage unit 34 stores information such as welding conditions and total welding time set by the operation of the operation unit 33. Welding conditions include, for example, the welding current and welding voltage recommended for the material and thickness of the work W, the shape of the groove, the type of joint, the posture of the work W, the moving speed of the predetermined welding torch 3, the posture and the angle, and the like. Information on set values such as the feed rate of the welded wire. The welding conditions set on the welding torch 3 side are the recommended welding current, welding current, welding wire feeding speed, etc. when the welding torch 3 is welded at a specific moving speed, posture, and angle. The content is not particularly limited as long as it is information that serves as a reference for setting. Further, the welding conditions may be set on the welding power supply 1 side.

The torch state detecting unit 35 includes a sensor for detecting the moving speed of the welding torch 3, the posture of the welding torch 3 with respect to the moving direction, the angle of the welding torch 3 with respect to the vertical direction, and the like. For example, the torch state detection unit 35 includes an acceleration sensor 351 and a gyro sensor 352 (see FIG. 6) described later. The detection of the moving speed, posture and angle of the torch will be described later.

The control unit 36 is a computer having a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit) or a multi-core CPU, a ROM (Read Only Memory), a RAM (Random Access Memory), an input / output interface and the like. The control unit 36 executes a predetermined process such as setting welding conditions according to the operation of the operation unit 33. The control unit 36 transmits welding condition information indicating the set welding conditions to the welding power supply 1 by the torch side communication unit 31. Further, the control unit 36 executes a process of calculating the moving speed, posture, angle, etc. of the welding torch 3 based on the detected value detected by the torch state detecting unit 35. The control unit 36 transmits the speed information, the attitude information, and the angle information obtained by calculation to the welding power supply 1 by the torch side communication unit 31. Information regarding the state of the welding torch 3 is continuously transmitted to the welding power source 1 during welding. Further, the control unit 36 can also control the display unit 32 to appropriately display the welding conditions, the state of the welding torch 3, the information read from the storage unit 34, and the like.

FIG. 3 is a block diagram showing a configuration example of the welding power supply 1 according to the first embodiment. The welding power supply 1 includes a main control unit 11, a power supply unit 12, a power supply side communication unit 13, a first storage unit 14, a second storage unit 15, a learning unit 17, and a welding state detection unit 16.

The main control unit 11 is a microcomputer having a CPU, and controls the operation of each component unit constituting the welding power supply 1.

The power supply unit 12 is a circuit that supplies electric power for performing arc welding to the welding torch 3. The power supply unit 12 converts the three-phase AC power input from the power system P into power suitable for arc welding and outputs it. Specifically, the power supply unit 12 outputs a required welding current and welding voltage based on the welding information given from the learning unit 17 via the main control unit 11. The welding information shows the welding voltage, welding current, welding wire feeding speed, etc., which can be adapted to specific welding conditions, the moving speed of the welding torch 3, the attitude and the angle, and obtain good welding results. Information. Further, the power supply unit 12 converts the three-phase AC power into electric power suitable for driving control of the wire feeding device 2 and the welding torch 3, and supplies the three-phase AC power to the wire feeding device 2 and the welding torch 3.

The power supply side communication unit 13 is a circuit that communicates with the wire feeding device 2 and the welding torch 3. The power supply side communication unit 13 modulates the signal given from the control unit 36 according to a predetermined communication protocol, and transmits the signal to the wire feeding device 2 and the welding power supply 1 through the signal line 8. For example, the power supply side communication unit 13 wire feeds a control signal instructing the welding wire to be fed at a required feeding speed based on the welding information given from the learning unit 17 via the main control unit 11. Send to device 2. Further, the signal transmitted from the welding torch 3 and the wire feeding device 2 is received, demodulated, and the demodulated signal is given to the main control unit 11, the learning unit 17, the welding state detection unit 16, and the like.

The first storage unit 14 is a non-volatile memory such as an EEPROM and a flash memory, and stores a program necessary for the operation of the main control unit 11.

The second storage unit 15 is a non-volatile memory similar to the first storage unit 14. The second storage unit 15 is a process of calculating welding information adapted to the moving speed, attitude and angle of the welding torch 3, a process of accumulating learning information 15b for learning a neural network for calculating welding information, and the neural. A computer program 15a for causing the learning unit 17, which is a computer, to execute a process of machine learning the network or the like is stored.

The welding state detecting unit 16 includes a current sensor that detects a welding current flowing through an arc, a welding torch 3, a voltage sensor that detects a voltage applied to the work W, and the like during the welding process. The welding state detecting unit 16 outputs welding state information indicating the welding state such as the detected welding current and welding voltage to the learning unit 17.

The learning unit 17 is a computer having a processor such as a CPU, GPU or multi-core CPU, a ROM, a RAM, an input / output interface, and the like. The learning unit 17 calculates welding information adapted to the set welding conditions, the moving speed of the welding torch 3, the posture and the angle, and outputs the calculated welding information to the main control unit 11. The learning unit 17 calculates welding information using a neural network as described later. The main control unit 11 supplies the welding information calculated by the learning unit 17 to the power supply unit 12 and controls the supply of the welding current and the welding voltage.

Further, the learning unit 17 has functional units such as a torch information acquisition unit 17a, a welding state information acquisition unit 17b, and an accumulation processing unit 17c, and performs a process of collecting and accumulating learning information 15b necessary for its own learning. Execute. The power supply side communication unit 13 communicates with the torch side communication unit 31 and receives welding condition information, speed information of the welding torch 3, attitude information and angle information, welding condition information, and the like. The torch information acquisition unit 17a acquires welding condition information, speed information, attitude information, and angle information of the welding torch 3 received by the power supply side communication unit 13 from the power supply side communication unit 13. The welding state information acquisition unit 17b acquires welding state information indicating welding voltage, welding current, etc. from the welding state detection unit 16 as information for determining the quality of the welding result. The storage processing unit 17c associates the welding information used for welding, each information acquired by the torch information acquisition unit 17a and the welding state information acquisition unit 17b, and stores them in the second storage unit 15 as learning information 15b.

Further, the second storage unit 15 stores the information related to the welding procedure manual as learning information 15b. Information related to the welding procedure manual is, for example, the material, thickness, groove shape, joint type, work W posture, recommended welding torch 3 movement speed, posture, angle, and welding. Information such as current, welding voltage, and feeding speed of welding wire, and the information is stored as learning information 15b.
The learning unit 17 can execute a process of machine learning itself based on the learning information 15b stored in the second storage unit 15. Machine learning enables the learning unit 17 to output welding information that gives better welding results. The details of the configuration related to the calculation and learning of welding information by the learning unit 17 will be described later.

FIG. 4 is an external view showing a configuration example of the welding torch 3. 4A is a front view of the welding torch 3, and FIG. 4B is a plan view of the welding torch 3. The welding torch 3 includes a torch body 37, a nozzle 371, a handle 38, a control board 381, a torch switch 331, an operation button 332, a display 321, an acceleration sensor 351 and a gyro sensor 352, and a torch cable 39.

The torch body 37 is a metal tubular member, and has a liner through which a welding wire is inserted, a first power cable 41, and a gas pipe 7 arranged therein. A nozzle 371 is provided at the tip of the torch body 37. The torch body 37 has a curved portion so that the welder can easily point the nozzle 371 at the work W.

The handle 38 is a portion for the welder to grip, and holds the base end portion of the torch body 37. The welding operator grips the handle 38 and performs welding work. The handle 38 is provided with a torch switch 331, an operation button 332, and a display 321. Further, a control board 381 is arranged inside the handle 38. The control board 381 is provided with circuits constituting a torch-side communication unit 31, a display unit 32, an operation unit 33, a storage unit 34, a torch state detection unit 35, and a control unit 36.

The torch switch 331 is an operating means for accepting operations for starting and stopping welding, and is provided at a position where a welding operator holding the handle 38 can easily push and operate with the index finger. When the torch switch 331 is turned on (pressed), an operation signal is output to the control unit 36, and the operation signal is input to the welding power source 1, so that the welding power source 1 outputs welding power. When the on operation is released, the welding power source 1 stops the output of the welding power. That is, welding is performed only while the torch switch 331 is pressed.

The display 321 is a panel for displaying various tables, and is provided on the opposite side of the torch switch 331 in the handle 38 so that the welding operator can easily see the screen.

The operation button 332 is an operation means for switching the screen, performing various settings, and changing the settings, and is arranged between the grip portion of the handle 38 and the display 321 on the same side as the display 321 of the handle 38. ing. The operation button 332 includes an upper button 332a, a lower button 332b, a left button 332c, and a right button 332d. When each button is pressed, the corresponding operation signal is output to the control unit 36, and the control unit 36 performs the corresponding processing.

The acceleration sensor 351 is a three-axis acceleration sensor 351 that detects acceleration in each axial direction and outputs the detected value to the control unit 36. The gyro sensor 352 is a 3-axis gyro sensor 352, detects the angular velocity around each axis, and outputs the detected value to the control unit 36. The control unit 36 calculates speed information indicating the speed of the welding torch 3 based on the detection values input from the acceleration sensor 351 and the gyro sensor 352 of the torch state detection unit 35. Further, the control unit 36 determines the attitude information indicating the posture of the welding torch 3 with respect to the moving direction of the welding torch 3 and the angle of the welding torch 3 with respect to the vertical direction based on the detection values input from the acceleration sensor 351 and the gyro sensor 352. Calculate the indicated angle information.

The method of calculating the moving speed of the welding torch 3 by the control unit 36 is not particularly limited. For example, the calculation may be performed only from the value detected by the gyro sensor 352 without using the acceleration sensor 351. Further, the calculation may be performed only from the value detected by the acceleration sensor 351 without using the gyro sensor 352.

<Movement speed of welding torch>
The recommended moving speed of the welding torch 3 in semi-automatic arc welding is determined according to the material of the work W, the material of the welding wire, the diameter, the feeding speed, and the like. The welding operator performs welding while moving the welding torch 3 so as to maintain the recommended speed. However, it is difficult to keep the welding torch 3 moving at the recommended speed, and in reality, the moving speed of the welding torch 3 becomes faster or slower than the recommended speed.

FIG. 5 is a schematic diagram showing the relationship between the moving speed and welding voltage of the welding torch 3 and the welding state, and an arc A is formed between the tip of the welding wire D protruding from the tip of the welding torch 3 and the work W. It shows the state of being generated and welding.

FIG. 5A shows a state in which the welding torch 3 is moved at the recommended speed. In this case, the thickness of the bead B formed on the work W is an appropriate thickness, and the length La of the arc A between the tip of the welding wire D and the surface of the bead B is also an appropriate length. It has become.

FIG. 5B shows a state in which the welding torch 3 is moved at a speed higher than the recommended speed. In this case, the thickness of the bead B formed on the work W becomes thinner than when the work W is moved at the recommended speed. As a result, the length La of the arc A between the welding wire D and the surface of the bead B becomes long, and spatter is likely to occur.

FIG. 5C shows a state in which the welding torch 3 is moved at a speed higher than the recommended speed as in FIG. 5B, and the welding voltage is lower than the specified set voltage. In this case, the thickness of the bead B is as thin as in FIG. 5B. However, by setting the welding voltage low, the welding wire D becomes difficult to melt, and the length L of the welding wire D protruding from the tip of the welding torch 3 is longer than that in FIGS. 5A and 5B. With this, the length La of the arc A between the tip of the welding wire D and the surface of the bead B becomes an appropriate length, and the generation of spatter is suppressed.

On the contrary, when the welding torch 3 is moved at a speed slower than the recommended speed, the thickness of the bead B becomes thick, so that the length La of the arc A between the tip of the welding wire D and the surface of the bead B becomes short. Become. In this case, by making the welding voltage higher than the specified set voltage, melting of the welding wire D is promoted, and the length L of the welding wire D protruding from the tip of the welding torch 3 becomes shorter than in the case of FIG. 5A. As a result, the length La of the arc A becomes an appropriate length.

In this way, by adjusting the welding voltage according to the moving speed of the welding torch 3, the length La of the arc A can be adjusted, and the deterioration of the welding quality due to the change in the moving speed of the welding torch 3 can be suppressed. can.

<Angle with respect to the moving direction of the welding torch>
FIG. 6 is a schematic view showing the posture of the welding torch 3 with respect to the moving direction. In the figure, the thick arrow indicates the moving direction of the welding torch 3.
In semi-automatic arc welding, the welding state changes depending on the posture of the welding torch 3 with respect to the moving direction. As shown in the right figure, the welding method in which the welding torch 3 is tilted in the moving direction is called a retreat method. In the case of the receding method, a bead with a deep penetration, a high surplus, and a narrow width is formed. As shown in the left figure, a welding method in which the welding torch 3 is tilted in the direction opposite to the moving direction is called a forward method. In the case of the forward method, a bead with a shallow penetration, a low surplus, and a wide bead is formed.

In semi-automatic arc welding, the recommended posture of the welding torch 3 is determined according to the material of the work W, the material of the welding wire, the diameter, the feeding speed, etc., as well as the moving speed. ing. The welding operator moves the welding torch 3 while maintaining the posture of the welding torch 3 to perform welding. However, it is difficult to keep the posture of the welding torch 3 constant. When the posture of the welding torch 3 changes, the welding state changes and the welding quality may deteriorate. In particular, it is known that when the inclination of the welding torch 3 becomes too large by the forward method, the generation of spatter increases.

“Within a predetermined range” shown in the central figure indicates a range recommended as the posture of the welding torch 3. When the posture of the welding torch 3 is within the range, good welding results can be obtained by performing welding using a specified reference voltage value as the welding voltage.
However, as shown in the left figure, when the welding torch 3 is tilted too much in the opposite direction of the moving direction (advance angle is exceeded), if a value obtained by adding a predetermined value α to the specified welding voltage set value is set, the welding result is obtained. It tends to be improved. Further, as shown in the right figure, when the welding torch 3 is tilted too much in the moving direction (backward angle is exceeded), setting a value obtained by subtracting a predetermined value β from the set value of the specified welding voltage improves the welding result. Tends to be.

<Angle of welding torch>
FIG. 7 is a schematic view showing the angle of the welding torch 3 with respect to the vertical direction. In semi-automatic arc welding, welding is usually performed with the tip of the welding torch 3 facing downward, but as shown in FIG. 7, the tip is laterally (90 °) or upward (180 °) depending on the welding location. ) May be welded. In this case, when welding is performed with the tip of the welding torch 3 facing laterally or upward, it tends to be better to set the welding current lower than when welding with the tip facing downward.
For example, when the tip of the welding torch 3 is oriented in the lateral direction, setting a value obtained by subtracting a predetermined value α from a predetermined setting value of the welding current tends to improve the welding result. Further, when the tip of the welding torch 3 faces upward, setting a value obtained by subtracting a larger predetermined value β from the set value of the specified welding current tends to improve the welding result.

<Learning Department>
FIG. 8 is a block diagram showing a configuration example of the learning unit 17. The learning unit 17 includes an acquisition unit 171, a welding information calculation unit 172, a learning information reading unit 173, a pass / fail determination unit 174, and a learning processing unit 176.

The acquisition unit 171 includes a torch speed information acquisition unit 171a, a torch attitude information acquisition unit 171b, a torch angle information acquisition unit 171c, and a welding condition information acquisition unit 171d. The power supply side communication unit 13 shown in FIG. 3 communicates with the torch side communication unit 31 and receives speed information, attitude information, angle information, welding condition information, etc. of the welding torch 3, and the torch of the learning unit 17. The speed information acquisition unit 171a acquires the speed information received by the power supply side communication unit 13. Similarly, the torch attitude information acquisition unit 171b and the torch angle information acquisition unit 171c acquire the attitude information and the angle information received by the power supply side communication unit 13. Further, the welding condition information acquisition unit 171d acquires the welding condition information received by the power supply side communication unit 13. The acquisition unit 171 outputs the acquired speed information, attitude information, angle information, and welding condition information to the welding information calculation unit 172.

When the speed information, attitude information and angle information of the welding torch 3 and the welding condition information are input, the welding information calculation unit 172 can obtain welding information adapted to the state and welding conditions of the torch, that is, good welding results. It is provided with a welding information calculation neural network (Neural Network) 172a that outputs information indicating the obtained welding voltage, welding current, welding wire feeding speed, and the like. Hereinafter, the welding information calculation neural network 172a is referred to as a welding information calculation NN172a. The weld information calculation NN172a is a trained neural network having an input layer, an intermediate layer and an output layer. The intermediate layer is a plurality of layers. The input layer comprises a plurality of neurons into which the velocity information, attitude information and angle information of the welding torch 3 and welding conditions are input. The output layer comprises a plurality of neurons corresponding to the magnitudes such as welding current, welding voltage and feed rate of welding wire. For example, it has a neuron corresponding to a welding current = γ and a plurality of neurons corresponding to a welding current = 2γ, 3γ .... The same applies to the welding voltage and the feeding speed. Each neuron outputs information indicating the desired probability of the welding current, welding voltage and feed rate. Based on the information output from the output layer of the welding information calculation NN172a, the welding information calculation unit 172 determines the most desirable, that is, the welding current, welding voltage, welding wire feeding speed, etc. corresponding to the neuron having the highest output. Welding information is determined, and the determined welding information is given to the main control unit 11 and the power supply unit 12. The power supply unit 12 controls the welding current and the welding voltage by using the welding information output from the learning unit 17. The main control unit 11 uses the welding information output from the learning unit 17 to transmit a control signal instructing the welding speed and the like of the welding wire to the wire feeding device 2.
The structure of the welding information calculation NN172a, such as the number of layers in the intermediate layer and the number of neurons in each layer, is not particularly limited.

The learning information reading unit 173 reads out the learning information 15b stored in the second storage unit 15, and outputs the read learning information 15b to the quality determination unit 174 and the learning processing unit 176. The learning information 15b includes, for example, welding information used for welding, speed information of the welding torch 3, attitude information and angle information, and information for determining the quality of the welding result. The information for determining the quality of the welding result is, for example, the welding voltage, the welding current, etc. detected during welding. Hereinafter, the welding voltage and welding current are referred to as welding monitor information. The learning information reading unit 173 outputs welding monitor information to the quality determination unit 174. Further, the learning information reading unit 173 outputs the speed information, the attitude information, the angle information, and the welding information of the welding torch 3 to the learning unit 17.

When the welding monitor information is input, the quality determination unit 174 includes a quality determination neural network 175 that outputs information indicating the quality of the welding result related to the welding process when the welding monitor information is obtained. Hereinafter, the pass / fail determination neural network 175 is referred to as a pass / fail determination NN175. The pass / fail determination NN175 is, for example, a learned recurrent neural network (RNN), and has an input layer, an intermediate layer, and an output layer.
The pass / fail judgment NN175 includes, for example, a first neuron in which a neuron to which welding monitor information which is time-series data is input is provided in the input layer and outputs information indicating a probability that the welding result is good, and a first neuron which outputs information indicating the probability that the welding result is good, and the welding result is poor. The output layer is provided with a second neuron that outputs information indicating a certain probability. In this case, the information indicating the quality is the information output from the first and second neurons.
Further, the pass / fail determination NN175 may be provided with a neuron that outputs the pass / fail of the welding result as a binary value in the output layer. In this case, the information indicating the quality is binary information output from the neuron.
Further, the pass / fail determination NN175 may include a neuron that outputs an analog value indicating the degree of pass / fail of the welding result in the output layer.
The quality determination NN175 uses the welding monitor information (input data) and the information indicating the quality of the welding result corresponding to the welding monitor information (teacher data) as the learning information of the quality determination NN175, and is a recursive deep neural before learning. It may be learned by giving it to the network.
The structure of the pass / fail judgment NN175, such as the number of layers in the intermediate layer and the number of neurons in each layer, is not particularly limited. Further, the pass / fail judgment NN175 does not necessarily have to be a recurrent neural network, and may be configured by other types of neural networks. Further, the quality determination unit 174 does not necessarily have to be a neural network, and may be configured to extract a feature amount from the waveform of the welding current or the welding voltage and determine the quality of the welding result. Further, based on the welding monitor information, the short circuit time, the frequency of arc breakage, and the like may be calculated to determine the quality of the welding result.

The learning processing unit 176 acquires welding conditions, speed information, attitude information, angle information, welding information, etc. of the welding torch 3 output from the learning information reading unit 173. Further, the learning processing unit 176 acquires information indicating the quality of the welding result output from the quality determination unit 174.

The learning method of welding information calculation NN172a will be described.
(1) Initial learning First, the initial learning of welding information calculation NN172a will be described. Initial learning is machine learning before using an arc welding system in the field. In the initial learning, a skilled person performs welding at the welding voltage, welding current, welding wire feeding speed, moving speed, posture and angle of the welding torch 3 recommended under specific welding conditions, and the learning information 15b is provided. accumulate. In addition, assuming that welding was performed at a speed faster or slower than the recommended moving speed of the welding torch 3, the welding voltage, welding current, etc. were adjusted so that good welding results could be obtained under the relevant conditions. Then, welding is performed and the learning information 15b is accumulated. Similarly, assuming that the posture and angle of the welding torch 3 deviate from the recommended posture and angle, adjust the welding voltage, welding current, etc. so that good welding results can be obtained under the conditions. Welding is performed and learning information 15b is accumulated.
In addition, the welding current and welding voltage are set to values deviating from the recommended values, and under the welding conditions, a skilled person tries welding so that good welding results can be obtained, and learns the information obtained by the welding. It may be stored as information 15b.
Further, standard welding conditions, welding voltage, welding current, welding wire feeding speed, welding torch 3 speed information, attitude information and angle information are directly stored in the second storage unit 15 as learning information 15b, and are actually stored. It is good to make up for the lack of learning information 15b obtained by welding by a skilled worker.
The learning processing unit 176 selects the learning information 15b for which a good welding result is obtained from the accumulated learning information 15b. That is, the learning information 15b representing the welding voltage, the welding current, the feeding speed of the welding wire, the moving speed of the welding torch 3, the attitude, and the angle at which good welding results can be obtained under specific welding conditions is selected. Then, the learning processing unit 176 trains the welding information calculation NN172a using the selected learning information 15b. The learning processing unit 176 inputs the speed information, attitude information, angle information, and welding condition information of the welding torch 3 into the welding information calculation NN172a, and acquires the welding information output from the welding information calculation NN172a. Then, the error between the output welding information and the welding information when a good welding result is obtained is calculated, and the weighting coefficient that characterizes the welding information calculation NN172a is used in the error back propagation method so that the error becomes small. Make corrections. By modifying the weighting coefficient, the welding information calculation NN172a can output welding information such as welding current and welding voltage so that better welding results can be obtained.

(2) Additional learning Next, the additional learning of the welding information calculation NN172a will be described. The additional learning is machine learning performed based on the learning information 15b obtained by using the arc welding system in the field. In other words, the additional learning is a process of correcting the weighting coefficient of the welding information calculation NN172a so as to cope with the welding conditions and the state of the welding torch 3 which were not assumed in the initial learning.
The welding power supply 1 stores information related to welding as learning information 15b each time a welding worker performs welding at the site. The welding information calculation NN172a executes the additional learning process when the welding operator performs the additional learning operation.
As in the initial learning, the learning processing unit 176 may select the learning information 15b for which a good welding result is obtained to learn the welding information calculation NN172a. That is, the welding information calculation NN172a is additionally learned so that the welding information can be calculated based on the welding conditions not assumed in the initial learning and the optimum welding information in the state of the welding torch 3. In this case, it is preferable to mix the learning information 15b used in the initial learning and the learning information 15b accumulated in the field to execute additional learning.

Further, the welding power supply 1 has an additional learning mode that automatically adjusts welding information so that good welding results can be obtained and supports the accumulation of learning information 15b necessary for additional learning of the welding information calculation NN172a. When the welding operator operates the operation unit 33 of the welding torch 3, the welding power supply 1 shifts to the additional learning mode. In the additional learning mode, the welding voltage and welding current during welding are monitored, and the welding current, welding voltage, etc. are adjusted in real time so that good welding results can be obtained. Hereinafter, pulse welding will be described as an example.
The learning unit 17 detects the ratio of the short circuit time to the welding time, the stability of the arc, the variation of the short circuit time, and the arc breakage during welding based on the welding voltage and the welding current acquired by the welding state information acquisition unit 17b. be able to. An indicator of arc stability is obtained by subtracting the detected weld voltage fluctuation cumulative value from the cumulative voltage fluctuation target value.

When the short-circuit time ratio is smaller than the reference value, the learning unit 17 corrects the welding information so that the welding current, the welding voltage, the pulse peak current, the pulse base current, and the pulse frequency are all large, and the learning unit 17 sends the welding information to the power supply unit 12. Output. On the contrary, when the ratio of the short circuit time is larger than the reference value, the welding information is corrected so that the welding current, the welding voltage, the pulse peak current, the pulse base current, and the pulse frequency are all reduced and output to the power supply unit 12. .. The learning unit 17 changes, for example, the welding current, the welding voltage, the pulse peak current, and the like in order, and when the welding state is improved, holds the changed values.
When the ratio of the short-circuit time deviates from the reference value, the learning unit 17 corrects the welding information by trial and error so that the short-circuit current, the rise and fall times of the pulse become large or small, and after the correction. Welding information is output to the power supply unit 12. If the welding result is not improved, the learning unit 17 corrects the welding information so that the short-circuit current, the rising edge of the pulse, and the falling edge time become the original set values.

Further, the learning unit 17 outputs a smaller welding current when the arc is unstable, a large variation in short circuit time, or an arc breakage during welding, and a larger welding voltage and short circuit current. Welding information is corrected so that is output, and the corrected welding information is output to the power supply unit 12.
The learning unit 17 has a pulse peak current, a pulse base current, a pulse frequency, and a pulse rise when the arc is unstable, the short circuit time varies widely, and the arc breaks during welding. Welding information is corrected by trial and error so that the fall time becomes large or small, and the corrected welding information is output to the power supply unit 12. If the welding result is not improved, the learning unit 17 corrects the welding information so as to return to the original set value.

In the additional learning mode, the learning unit 17 stores the corrected welding information, the speed information of the welding torch 3, the attitude information and the angle information, the welding condition information, etc. in the second storage unit 15 as learning information 15b. ..

By executing welding in the additional learning mode, it is possible to accumulate optimum welding information under the welding conditions and the state of the welding torch 3 which were not assumed in the initial learning. Then, the learning unit 17 additionally learns the welding information calculation NN172a by the same procedure as the initial learning.

In the arc welding system according to the first embodiment, the optimum welding current, welding voltage, and welding wire adapted to changes in the moving speed, posture, angle, etc. of the welding torch 3 are used without relying on the skill level of the welding operator. Welding information indicating the feeding speed and the like can be calculated.

Further, the welding operator only operates the operation unit 33 of the welding torch 3 and selects the welding conditions, and the learning unit 17 of the welding power source 1 operates the actual moving speed of the welding torch 3 under the selected welding conditions. Welding information such as the optimum welding current adapted to changes in posture, angle, etc. can be calculated.

Further, the welding power supply 1 can collect and store learning information 15b for machine learning the welding information calculation unit 172, and can additionally learn welding information calculation NN172a using the accumulated learning information 15b. can. Further, the welding power supply 1 can additionally learn the welding information calculation NN172a so as to automatically determine the quality of the welding result and output the welding information in which a good welding result can be obtained.

In the first embodiment, an example in which the welding information calculation unit 172 outputs welding information based on the state and welding conditions of the welding torch 3 has been described, but based on the moving speed, posture, and angle of the welding torch 3. , The welding voltage, the welding current, the feeding speed of the welding wire, and the like may be configured to be calculated and output as a correction amount. If the correction amount does not greatly depend on the welding conditions, the welding information calculation unit 172 may be configured to calculate the correction amount based only on the moving speed, posture, and angle of the welding torch 3.

Further, although the example in which the second storage unit 15 and the learning unit 17 are provided in the welding power supply 1 has been described, the second storage unit 15 and the learning unit 17 may be provided in the external server device. The welding power supply 1 transmits the learning information 15b to the server device. The welding information calculation unit 172 is machine-learned on the server device side, and the parameters defining the welding information calculation unit 172 after learning are transmitted to the welding power supply 1. The server device may be configured to collect and store learning information 15b from a plurality of welding power sources 1. The welding power supply 1 receives the parameters transmitted from the server device, and reconstructs the welding information calculation unit 172 using the received parameters.

Further, in the first embodiment, the device and the welding system for performing arc welding have been described as an example, but the present embodiment 1 is also applied to other types of welding such as laser welding, electron beam welding, and gas welding. Can be done. The same applies to other embodiments described later.

(Embodiment 2)
FIG. 9 is a block diagram showing a configuration example of the welding torch 203 according to the second embodiment. The arc welding system according to the second embodiment is different from the first embodiment in that the welding information calculation NN234a is provided on the welding torch 203. Therefore, the above differences will be mainly described below. Since other configurations and actions and effects are the same as those in the embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The welding torch 203 according to the second embodiment includes a welding information calculation unit 234 having a learned welding information calculation NN234a. The configuration of the welding information calculation NN234a is the same as that of the welding information calculation NN172a provided in the welding power supply 1 of the first embodiment.

However, the additional learning process of the welding information calculation NN234a may be executed by the welding power supply 1 or the server device having the same function as the learning unit 17 of the first embodiment. The welding torch 203 acquires and stores the parameters of the neural network obtained as a result of the additional learning.

The welding torch 203 according to the second embodiment calculates welding information adapted to the welding conditions and the state of the welding torch 203 by the welding information calculation NN234a, and outputs the calculated welding information to the welding power supply 1. By supplying the welding information calculated by the welding information calculation unit 234 to the welding power source 1, good welding results can be obtained regardless of changes in the moving speed, posture and angle of the welding torch 203.

Since the welding torch 203 is configured to calculate the welding information and transmit the calculated welding information to the welding power source 1, the amount of information transmitted from the welding torch 203 to the welding power source 1 can be reduced. Therefore, the welding power supply 1 can supply the optimum welding voltage and welding current to the welding torch 203 without delaying the state change of the welding torch 203.

(Embodiment 3)
FIG. 10 is a block diagram showing a configuration example of the welding torch 303 according to the third embodiment. The arc welding system according to the third embodiment uses the welding voltage and welding current detected during welding, image information obtained by imaging the welding marks, and welding sound information obtained by collecting sound during welding. Since the difference from the first embodiment is that the quality of the welding result is determined and the welding information calculation unit 172 is machine-learned, the above differences will be mainly described below. Since other configurations and actions and effects are the same as those in the embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The welding torch 303 according to the third embodiment further includes an image pickup unit 337 and a microphone 338. The image pickup unit 337 is a camera that captures a welding mark. The microphone 338 collects welding sounds during welding. The image information obtained by imaging and the audio information obtained by collecting sound are transmitted to the welding power supply 1 as welding state information. The welding power supply 1 receives welding monitor information, voice information, and image information at the power supply side communication unit 13, and gives each received information to the learning unit 17.

FIG. 11 is a block diagram showing a configuration example of the quality determination unit 174 according to the third embodiment. The pass / fail determination unit 174 includes a first pass / fail determination unit 174a, a second pass / fail determination unit 174b, a third pass / fail determination unit 174c, and a comprehensive evaluation unit 175d.
The first quality determination unit 174a has a quality determination RNN (Recurrent Neural Network) 175a that outputs information for determining the quality of the welding result based on the input time-series welding monitor information.
The second quality determination unit 174b has a quality determination RNN175b that outputs information for determining the quality of the welding result based on the input time-series voice information.
The third quality determination unit 174c has a quality determination CNN (Convolutional Neural Network) 175c that outputs information for determining the quality of the welding result based on the input image information.
The information output from the first pass / fail determination unit 174a, the second pass / fail determination unit 174b, and the third pass / fail determination unit 174c is, for example, information indicating the probability that the welding result is good.

The comprehensive evaluation unit 175d determines the quality of the welding result based on the information output from the first quality determination unit 174a, the second quality determination unit 174b, and the third quality determination unit 174c, and learns the quality determination result. Output to the processing unit 176. For example, when all of the information output from each determination unit is equal to or greater than the threshold value, the comprehensive evaluation unit 175d may determine that the welding result is good. The pass / fail determination method is an example, and the pass / fail may be determined by calculating a statistic such as an average value based on each information and comparing the statistic with a threshold value.

According to the arc welding system according to the second embodiment, in addition to the welding monitor information of the welding current and the welding voltage detected during welding, the welding sound, the image obtained by imaging the welding marks, and the like are added to make it more accurate. It is possible to determine the quality of the welding result and output the welding information to improve the welding result.

In the additional learning mode in the first embodiment, an example in which the welding current and the welding voltage are adjusted based on the ratio of the short circuit time, the variation of the short circuit time, etc., and the welding information calculation unit 172 is additionally learned has been described. Welding current and welding voltage can also be adjusted based on the information and image information. The learning unit 17 can specify the degree of penetration, the bead width, the uniformity of the bead width, the spatter, and the like based on the image information.

For example, the learning unit 17 corrects the welding information so that the welding current, the welding voltage, the pulse peak current, the pulse base current, and the pulse frequency are all large when the penetration is small or the bead width is narrow, and the power supply is supplied. Output to unit 12. On the contrary, when the melt-off occurs or the bead width is wide, the welding information is corrected so that the welding current, the welding voltage, the pulse peak current, the pulse base current, and the pulse frequency are all reduced, and the power supply unit 12 Output to.
Further, the learning unit 17 corrects the welding information so that a smaller welding current and a short-circuit current are output and a larger welding voltage is output when there is a lot of spatter and the bead width is not uniform. Welding information is output to the power supply unit 12.

Further, the learning unit 17 weights and adds the excess / deficiency amount of penetration, the target value of the bead width, the difference between the actual bead width, the spatter amount, and the index of arc stability, thereby welding current or welding. It may be configured to calculate the correction amount of the voltage.

(Embodiment 4)
FIG. 12 is a block diagram showing a configuration example of the learning unit 17 according to the fourth embodiment. Since the arc welding system according to the fourth embodiment is different from the first embodiment in that it includes a plurality of welding information calculation NN172a different for each welding condition, the above differences will be mainly described below. Since other configurations and actions and effects are the same as those in the embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The welding information calculation unit 172 according to the second embodiment includes a plurality of welding information calculation NN172a different for each welding condition. For example, a plurality of welding information calculation NN172a different for each of the material, thickness, groove shape, joint type, and posture of the work W are provided. The learning unit 17 stores the correspondence between the plurality of welding information calculation NN172a and the plurality of welding conditions.

The welding operator selects welding conditions by operating the operation unit 33 of the welding torch 3. The selected welding condition information is transmitted to the welding power source 1, and the welding power source 1 receives the welding condition information. The learning unit 17 selects the welding information calculation NN172a corresponding to the welding condition information received by the welding power supply 1, and calculates the welding information using the selected welding condition calculation NN.

According to the arc welding system according to the fourth embodiment, since the welding information calculation NN172a different for each welding condition is provided, the welding information calculation unit 172 can be efficiently machine-learned.

(Embodiment 5)
FIG. 13 is a block diagram showing a configuration example of the learning unit 17 according to the fifth embodiment. Since the arc welding system according to the fourth embodiment is different from the first embodiment in that it includes an operation prediction unit 172b, the above differences will be mainly described below. Since other configurations and actions and effects are the same as those in the embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The welding information calculation unit 172 according to the fifth embodiment includes an operation prediction unit 172b that predicts the operation of the welding torch 3. The motion prediction unit 172b is a recurrent neural network that predicts the motion of the welding torch 3 based on, for example, temporal changes in the moving speed, posture, and angle of the welding torch 3. The neural network has an input layer having neurons for inputting speed information, attitude information and angle information of the welding torch 3, an intermediate layer, and an output having neurons for outputting predicted speed information, attitude information and angle information. Has a layer. The learning processing unit 176 so that the difference between the speed information, attitude information and angle information of the welding torch 3 output from the motion prediction unit 172b and the actual speed information, attitude information and angle information of the welding torch 3 becomes small. , Continuously machine-learn the motion prediction unit 172b. The welding information calculation unit 172 calculates welding information based on the velocity information, attitude information, and angle information of the welding torch 3 predicted by the motion prediction unit 172b.

According to the arc welding system according to the fifth embodiment, the operation of the welding torch 3 can be predicted and welding information adapted to the predicted operation of the welding torch 3 can be calculated, so that better welding results can be obtained. Welding information can be output.

(Embodiment 6)
FIG. 14 is a block diagram showing a configuration example of the learning unit 17 according to the sixth embodiment. The arc welding system according to the sixth embodiment is different from the fifth embodiment in that it includes a plurality of motion prediction units 172b. Hereinafter, the above differences will be mainly described. Since other configurations and actions and effects are the same as those in the embodiment, the corresponding parts are designated by the same reference numerals and detailed description thereof will be omitted.

The welding information calculation unit 172 according to the sixth embodiment includes a plurality of operation prediction units 172b. Each motion prediction unit 172b has, for example, a recurrent neural network similar to that of the fifth embodiment in which the habits of a plurality of welders are learned. The welding operator selects the motion prediction unit 172b that has learned his / her own habit by operating the operation unit 33 of the welding torch 3. The welding information calculation unit 172 calculates welding information using the selected motion prediction unit 172b.

According to the arc welding system according to the sixth embodiment, the welding operator can select the motion prediction unit 172b, which is a neural network that has learned his / her own habit. The learning unit 17 can predict the movement of the welding torch 3 that differs for each welding operator using the selected motion prediction unit 172b, and can output welding information that gives good welding results.

The embodiments disclosed this time should be considered to be exemplary in all respects and not restrictive. The scope of the present invention is indicated by the scope of claims, not the above-mentioned meaning, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims.

1 Welding power supply 2 Wire feeder 3,203,303 Welding torch 5 Power transmission line 6 Gas cylinder 7 Gas piping 8 Signal line 11 Main control unit 12 Power supply unit 13 Power supply side communication unit 14 1st storage unit 15 2nd storage unit 15a Computer program 15b Learning information 16 Welding state detection unit 17 Learning unit 17a Torch information acquisition unit 17b Welding status information acquisition unit 17c Accumulation processing unit 31 Torch side communication unit 32 Display unit 321 Display 33 Operation unit 331 Torch switch 332 Operation button 34 Storage Part 35 Torch state detection part 351 Acceleration sensor 352 Gyro sensor 36 Control part 37 Torch body 337 Imaging part 338 Microphone 371 Nozzle 38 Handle 381 Control board 39 Torch cable 41 1st power cable 42 2nd power cable 171a Torch speed information acquisition part 171b Torch attitude information acquisition unit 171c Torch angle information acquisition unit 171d Welding condition information acquisition unit 172,234 Welding information calculation unit 172a, 234a Welding information calculation NN
172b Operation prediction unit 173 Learning information reading unit 174 Good / bad judgment unit 174a First good / bad judgment unit 174b Second good / bad judgment unit 174c Third good / bad judgment unit 175a, 175b Good / bad judgment RNN
175c Pass / Fail Judgment CNN
175d Comprehensive Evaluation Department 175 Good / Bad Judgment NN
176 Learning processing unit P power system W work

Claims (8)

Acquisition to acquire at least one of speed information indicating the speed of the welding torch that performs semi-automatic arc welding, attitude information indicating the posture of the welding torch with respect to the moving direction, and angle information indicating the angle of the welding torch with respect to the vertical direction. Department and
Welding information including welding current or welding voltage for performing welding adapted to the moving speed, attitude and angle of the welding torch based on at least one of speed information, attitude information and angle information acquired by the acquisition unit. Equipped with a trained welding information calculation unit to calculate
The welding information calculation unit is
It has a prediction unit that predicts the moving speed, attitude or angle of the welding torch based on at least one time change of speed information, attitude information and angle information acquired by the acquisition unit.
A welding information calculation device that calculates welding information adapted to the state of the welding torch based on at least one of the movement speed, the posture information indicating the posture and the angle, the posture information, and the angle information predicted by the prediction unit. A welding condition information acquisition unit for acquiring welding condition information indicating welding conditions that are a reference for welding information at a predetermined speed, posture, and angle of the welding torch is provided.
The welding information calculation unit is
Welding adapted to the state and welding conditions of the welding torch based on at least one of the speed information, attitude information and angle information acquired by the acquisition unit and the welding condition information acquired by the welding condition information acquisition unit. Information is calculated ,
Further, the welding information calculation unit is
It is provided with a neural network trained to output welding information adapted to the moving speed, posture or angle of the welding torch when at least one of speed information, posture information and angle information is input.
The welding information calculation device according to claim 1 . The welding information calculation unit is
Equipped with multiple different neural networks according to welding conditions
The welding information calculation device according to claim 2 . A welding condition information acquisition unit for acquiring welding condition information indicating welding conditions that are a reference for welding information at a predetermined speed, posture, and angle of the welding torch is provided.
The welding information calculation unit is
Welding adapted to the state and welding conditions of the welding torch based on at least one of the speed information, attitude information and angle information acquired by the acquisition unit and the welding condition information acquired by the welding condition information acquisition unit. Information is calculated,
Further, the welding information calculation unit is
It is provided with a neural network trained to output welding information adapted to the moving speed, posture or angle of the welding torch when at least one of speed information, attitude information and angle information and welding condition information are input.
The welding information calculation device according to claim 1 . The prediction unit
When at least one of speed information, attitude information, and angle information is input in time series, a plurality of speed information indicating the predicted movement speed, attitude, or angle of the welding torch, attitude information, or angle information is output. Neural network and
It is equipped with a reception unit that accepts the selection of the neural network.
The prediction unit
Using the neural network received by the reception unit, the moving speed, posture or angle of the welding torch is predicted.
The welding information calculation device according to any one of claims 1 to 4 . A welding torch provided with the welding information calculation device according to any one of claims 1 to 5 , which is supplied with electric power from a welding power source to perform welding.
The welding information calculation device is
A welding torch provided on the welding torch and including a transmission unit that transmits welding information calculated by the welding information calculation unit to the welding power supply. A welding power source provided with the welding information calculation device according to any one of claims 1 to 5 and supplying electric power to the welding torch.
The acquisition unit
At least one of the speed information, the attitude information and the angle information transmitted from the welding torch is acquired .
A welding power source that supplies electric power based on the welding information calculated by the welding information calculation unit. The welding information calculation device according to any one of claims 1 to 5 .
A welding power supply that supplies power to the welding torch,
A welding system including a welding torch to which electric power is supplied from the welding power source to perform welding.
The acquisition unit
At least one of the speed information, the attitude information and the angle information transmitted from the welding torch is acquired.
The welding power supply is
A welding system that supplies electric power based on the welding information calculated by the welding information calculation unit.

Images